RE: Kpps vs. Bandwidth [7:76249] posted 09/26/2003
- Subject: RE: Kpps vs. Bandwidth [7:76249]
- From: "Howard C. Berkowitz" <hcb@xxxxxxxxxxxx>
- Date: Fri, 26 Sep 2003 18:26:29 GMT
At 6:20 PM +0000 9/26/03, Priscilla Oppenheimer wrote:
>From what I understand, the PPS ratings are based on the bandwidth
>(capacity) of the interfaces that can be added to a router. Marketing
>engineers (yes, an oxymoron), load up the routers with high-speed interfaces
>and start whipping through the smallest possible packets. That way they get
>high PPS numbers.
http://www.isi.edu/in-notes/rfc2544.txt is a good start on how this
really should be done.
I would note that this deals with forwarding speed; control plane
(e.g., convergence) is a completely different problem.
>The number of packets that they can send through an interface depends on the
>bandwidth (capacity, speed) of that interface and other things like Zsombor
>mentioned such as Ethernet preambles and inter-frame gaps, (which don't
>apply to WAN links, though).
>So, the question is sort of backwards. The PPS number comes from a
>fully-loaded router with high-bandwidth connections. During development,
>Cisco does similar tests to determine what "fully loaded" will mean.
>To come up with a better answer regarding the amount of traffic a router can
>handle, which I think was the gist of your question, you would need to
>analyze the traffic volume and arrival rates, the buffering behavior of the
>router, the amount of RAM in the router, etc. Then we start getting into the
>realm of that other discussion going on about uilitzation and bandwidth. :-)
>Bandwidth just means capacity. What something is capable of may not match
>what it really does. Throughput, which is what you really care about, is a
>measurement of actual data that can be sent per timeframe. Since packet
>sizes vary, and aren't all the tiny ones used in the marketing tests, you
>need to do your own measurements with your own traffic, at the time of day
>where it matters, etc.
>Evan You wrote:
>> I know how many packets per second that each router can handle.
>> How do you
>> convert that to how much bandwidth that a router can handle.
>> For example, I
>> have a 2622 that is running at full E1/1984K (average is 80%
>> utilization and
>> it peaks often) and the CPU is at 25% to 30%. I know that this
>> router can
>> handle 15Kpps. So how does Kpps translate to bandwidth? I read
>> in another
> > thread that a DS-3 is equivalent to 150Kpps.
Message Posted at:
**Please support GroupStudy by purchasing from the GroupStudy Store:
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html